The study area encompasses the southern coastal region of Collier County, FL, USA centering on the Ten Thousand Islands (TTI) region, extending from the northwest portion of Everglades National Park westward to Marco Island (Fig. 1). The extent of the study area was dictated in part by manatee movements documented for radio-tagged individuals for this research. This area is a shallow, subtropical estuarine system consisting of tidally influenced rivers which discharge into inland bays enclosed by numerous mangrove islands. The coastal margin is fronted by shallow flats of the Gulf of Mexico. Tidal range is classified as microtidal (<1 m; Browder et al. 1986). Public-owned areas used by manatees include the Ten Thousand Islands National Wildlife Refuge, Everglades National Park, Rookery Bay National Estuarine Research Reserve, Collier-Seminole State Park, Fakahatchee Strand Preserve State Park, and Big Cypress National Preserve.
Two PTRs in the central portion of the study area were a special focus of this research: POI, at the north end of the Faka Union canal, and Big Cypress National Preserve (BCNP) headquarters (Fig. 1), which consists of canals that receive freshwater inflow from the upstream Big Cypress basin, and tidal inflow downstream from Halfway Creek. The area around POI is heavily influenced by the 77-km Faka Union canal system which was completed in 1971 to drain a failed real estate development, now slated for restoration (U.S. Army Corp of Engineers and South Florida Water Management District 2004; Swain and Decker 2009). The canal system intercepts sheet flow and rapidly diverts large water volumes as point discharge out the Faka Union Canal and into the Gulf of Mexico. Weirs were built to prevent over drainage during the dry season, but the canals have resulted in a general lowering of the water table, reduced groundwater seepage, and a shortened hydro-period (U.S. Army Corp of Engineers and South Florida Water Management District 2004). The southernmost weir restricts tidal incursion to the lower 7 km of the Faka Union canal (except during extreme tides or storm events).
Manatee Habitat Use
Although POI has long been known as an important manatee winter aggregation site (U.S. Fish and Wildlife Service 2001), we used aerial surveys and satellite telemetry to provide greater detail of current manatee habitat use within the region. To analyze the data, we divided the landscape into habitat patches and assigned each patch to one of four general zones for GIS analyses of manatee habitat use and movement. These zones were: (1) inland (rivers, canals, and basins); (2) enclosed bays; (3) inter-island corridors (channels and open areas between mangrove islands); and (4) offshore (shallow, near shore Gulf of Mexico; Fig. 2).
To identify areas of heavy use by manatees during winter, synoptic aerial survey data (Ackerman 1995) were obtained for the study region from 1991 to 2007 (FWRI 2009; four incomplete surveys were deleted). These surveys were flown as part of a statewide program using a standardized protocol and flight line that includes all of the major river systems, inland bays, and shallow Gulf shoals in the study area.
The aerial survey point data were overlaid with the habitat patches, and the results were summarized by zone and major winter aggregation sites. The summary data provide minimum counts of animals at surveyed sites, and underestimate the true population size by an unknown amount.
Winter telemetry data were available for 22 manatees captured and tagged at POI between 2001 and 2006 and provided data at a finer temporal and spatial resolution than aerial surveys. Data were obtained from floating Argos platform transmitter terminals (PTT; Telonics, Inc.) or GPS units (Lotek Wireless, Inc. or Telonics, Inc.) attached to manatees by flexible tethers to standardized peduncle belts (Reid et al. 1995). Argos location fixes were calculated from transmissions received by polar-orbiting satellites during four programmed duty cycles daily, potentially providing four standard quality locations per day. GPS location data were collected at time intervals ranging from 15 to 60 min.
Telemetry data were reduced to winter months only (November through March). Locational accuracy for the Argos data has a known variability, and only the two most accurate location classes (LC3, LC2) were kept: Service Argos, Inc. (1996) estimates that 68% of locations should lay within 150 m for LC3 and within 350 m for LC2. Data were further reduced to one location class per programmed duty cycle (Deutsch et al. 2003). The GPS data were subsampled to one location per 6-h interval approximating the Argos satellite fix intervals, and all winter location data from both tag types were merged for analysis.
GIS overlay procedures (ArcGIS identity function, E.S.R.I. 1997) were used to assign telemetry locations to one of the four habitat zones, river systems or canals, bays, inter-island corridors, and offshore (Fig. 2). We interpolated hourly Gulf temperatures to match date and time for each telemetry fix. The proportional manatee use of each habitat zone at different Gulf temperatures was calculated at 1°C intervals from 14°C to 27°C.
To investigate whether tagged manatees showed preferential use of different habitat zones based on water temperature, we calculated kernel density surfaces for each animal–winter combination using Gulf temperature to segregate the telemetry locations into three Gulf temperature bins: less than 18°C, 18°C to less than 20°C, and 20°C or above. We used the Home Range Extension for ArcView (Rodgers and Carr 1998) with a fixed kernel (500 m) for the utilization distribution (Worton 1989) and least square cross validation for the smoothing parameter (Silverman 1986). We summed the resulting surface grid files for all animals tracked for each threshold temperature bin, treating each animal–winter–temperature bin combination as a single sample. Maps were displayed using the 50th and tenth percentile contour levels to show moderately and heavily used areas.
Hydrodynamic Thermal Properties
Collection of Environmental Data
To investigate differences in thermal properties among the different water bodies frequented by manatees, data on water and air temperature, rainfall, salinity, flow, and stage in different habitat zones around the study area were obtained from permanent or temporary monitoring stations from several agencies (see Acknowledgments). Near-bottom temperatures were measured at all stations; surface temperatures were only available for the PTRs at POI and BCNP.
The PTRs were continuously monitored by USGS during four winters to characterize the vertical temperature regime. Temperature dataloggers (HOBO Water Temp v2 and Tidbit, Onset Computer Co., Pocasset, MA, USA) at near-surface and near-bottom depths recorded temperatures at 15 min intervals during two winters (2004/2005–2005/2006). These temperature-only probes were replaced by four YSI 600 XLM water quality sondes (YSI Incorporated, 1700 Brannum Lane, Yellow Springs, OH 45387-1107; www.ysi.com). Two probes were deployed at each site, within 15 cm of surface using a floating mount, and within 15 cm of the bottom. Date/time, temperature, conductance, and salinity (calculated using temperature and conductance) were logged every 30 min in subsequent winters (2006/2007–2007/2008).
Comparison of Thermal Properties of Gulf, Bay, and PTR Water Bodies
Hourly temperature data for seven site-depth locations (Fig. 1; POI top/bottom, BCNP top/bottom, Faka Union Bay, Fakahatchee Bay, Naples) representing key water bodies (PTR top/bottom, bay, and Gulf) for four winters (2004/2005–2007/2008) were analyzed to test for differences in water temperature as they related to temperature thresholds. We used the Kruskal–Wallis test followed by the Nemenyi–Damico–Wolfe–Dunn test (joint ranking; Hollander and Wolfe 1999) as implemented in the COIN library in R (Hothorn et al. 2008). To reduce the influence of warm periods on the results, the data were filtered to include only those periods when one or more sites fell below the 20°C manatee temperature threshold. This truncation resulted in strongly left-skewed distributions. Data were further filtered to obtain equal sample sizes among sites by rejecting any date–time where data were missing for one or more stations. To compare annual variation in winter water temperatures among the seven site-depth locations, we calculated the number of hourly readings at 1°C increments to provide an index of exposure time to different temperatures.
Analysis of Temperature and Salinity Profiles at the PTRs
To portray relationships among temperature and salinity, data for the two PTRs were plotted in vertical panels for the 2 years with salinity data. Mean daily discharge over the Faka Union weir also was plotted for POI (data were unavailable for BCNP).
Vertical Density Profiles at the PTRs
We calculated in-situ densities for surface and bottom layers at POI using the “International Equation of State of Water” (Fofonoff and Millard 1983) based on temperature, salinity, and pressure. These density calculations were used to examine whether density gradients associated with salinity stratification could offset opposing density gradients associated with temperature inversions. Changes in density result in fall turnover in lakes due to surface-water cooling (Reid and Wood 1976), and density gradients are commonly used to index the role of stratification in resisting turnover (Dyer 1997). For each bottom-surface measurement, we calculated and plotted the vertical difference in temperature and salinity for each paired sample. We overlaid this plot with a contoured surface of density differences to show the relationship between the vertical differences in density associated with vertical differences in salinity and temperature. The contoured surface of vertical density differences was generated as a least squares trend surface (six degree polynomial) using the “MASS” library in R (Venables and Ripley 2002).
Correlation analyses were conducted for data at POI to examine relationships among three key factors: (1) temperature difference (bottom–surface), (2) salinity difference (bottom–surface), and (3) freshwater discharge at the POI weir. The data were filtered to periods when the surface temperature fell below the threshold values of 18°C or 20°C.
Numerical Hydrology Modeling
To examine whether mechanisms associated with salinity and temperature stratification identified by analysis of the monitoring data could produce the observed patterns in water temperatures, we simulated conditions at POI with and without salinity stratification using a three-dimensional hydrodynamic model application of the Environmental Fluid Dynamics Code (EFDC) from the Virginia Institute of Marine Science (Hamrick 1992; U.S. Environmental Protection Agency 2002). This application of the EFDC model at POI is part of a larger project to investigate different restoration scenarios; here, we were only interested in comparing the POI system in the presence or absence of salt stratification. Because the physics-based EFDC application simulates the transport of variable density fluid in three dimensions, and explicitly models temperature and salinity, we viewed the model as an independent means of evaluating the patterns and mechanisms identified from the monitoring data at POI.
The three-dimensional application used a curvilinear grid with 104 columns and 50 rows (1,750 active cells) representing POI (Fig. 3). Grid size varied within the approximate range of 10–20 m with an average measurement of approximately 15 m. The northern boundary of active cells had a no-flow boundary representing the tide-blocking weir just north of POI. Volumetric sources were placed within the upper layers of the northern most cells to represent flow over the weir and into the canal. Boundaries for the northern flow volumes overtopping the weir were taken from South Florida Water Management District’s Database, DBHYDRO. The southern boundary was a water surface elevation boundary positioned approximately 130 m south of the port exit. The southern surface boundary water levels and the northern and southern salinity and temperature values were taken from a two-dimension surface-water/groundwater model of the TTI area (Swain and Decker 2009). Model bathymetry was determined by acoustic surveys and was defined by mean depth values.
General model input parameters, such as cell friction, heat transfer and evaporation coefficients, and vertical and horizontal diffusion, were initially determined from standard values and previous EFDC model applications and then adjusted for better model fit. The model requires extensive atmospheric data, including rain volumes, wind speed and direction, solar radiation, and air temperature, which was obtained from the Florida Automated Weather Network. Five vertical layers were used for the majority of testing and calibration and gave adequate representation of the vertical stratification of salinity and temperature.
Two simulations were run with and without the density effects of salinity variations over a 242-day time period (1 September 2004 to 30 April 2005). The salinity stratification used was relatively small compared to observed conditions, with the surface and bottom layers differing by about 8 psu. More recent periods could not be simulated due to the unavailability of key field measurements more recent than 2005.